All times are in CT (local time in St. Louis)
Tuesday, 18 November 2025
11:00 AM - 11:30 AM
TSUBAME4.0: More of Everyone's Supercomputer toward HPC-AI Era in Science Tokyo
Toshio Endo
Supercomputing Research Center, IIR, Institute of Science Tokyo
TSUBAME4.0 is a supercomputer in Institute of Science Tokyo (ex Tokyo Tech), equipped with 960 GPUs. This talk presents overview and current status of TSUBAME4.0 with Exascale AI performance.
3:15 PM - 3:45 PM
Designing for Responsiveness: Case Studies in Interactive High-Performance Systems
Hiroki Ohtsuji
Fujitsu Limited
TBA
Wednesday, 19 November 2025
12:00 PM - 12:30 PM
Strategy for GPU Application Development in Japan -Fugaku-NEXT and more-
Taisuke Boku
Advanced HPC-AI Research and Development Support Center (HAIRDESC) / Center for Computational Sciences, University of Tsukuba
TBA
3:00 PM - 3:30 PM
LLview: A Reporting Platform for the Exascale Era
Wolfgang Frings
Jülich Supercomputing Centre (JSC)
Diagnosing and reporting operational issues on large-scale HPC systems is a complex task that requires efficient monitoring and analysis.
At the Jülich Supercomputing Centre (JSC), we address this challenge with LLview, an open-source system and job reporting platform that supports both users and operators.
LLview provides near real-time metrics and analysis through an intuitive, role-based web interface, helping to optimize resource usage and overall system performance.
In this presentation, I will highlight LLview’s key features and showcase real examples where it has helped users and support teams identify issues and improve service efficiency.
Designed to scale with next-generation systems, LLview is ready to meet the reporting demands of the exascale era.
4:00 PM - 4:30 PM
From GR-MHD to GR-RMHD on GPUs: Faster Physics for Black-Hole Accretion Disks
Ryohei Kobayashi
Supercomputing Research Center, IIR, Institute of Science Tokyo
Black hole accretion disks serve as the central engines driving high-energy phenomena such as X-ray binaries, active galactic nuclei, and gamma-ray bursts, giving rise to complex physical processes such as turbulence induced by magnetorotational instability, intense heating, and relativistic jet formation. Traditional one-dimensional models and general relativistic magnetohydrodynamic (GR-MHD) simulations have captured many aspects of these systems but fail to reproduce radiative cooling and radiation pressure, which are crucial in high-accretion environments. To overcome these limitations, general relativistic radiation magnetohydrodynamics (GR-RMHD) simulations incorporate the interplay between radiation and plasma dynamics. However, including radiative transfer greatly increases computational cost, requiring advanced strategies for high-fidelity multidimensional simulations. In this talk, I will present a GR-RMHD code implemented in CUDA Fortran and MPI that achieves up to 39-fold speedup over CPU implementations with nearly linear GPU scalability. This acceleration enables detailed multidimensional studies of disks and jets, offering new insights into radiation cooling, pressure, and energy generation in compact astrophysical systems.